Proceedings of the 2015 IEEE 9th International Conference on Semantic Computing (IEEE ICSC 2015)

Semantic Data Mining: A Survey of

Ontology-based Approaches
Dejing Dou

Hao Wang

Haishan Liu

Computer and Information Science

Computer and Information Science

Computer and Information Science

University of Oregon

University of Oregon

University of Oregon

Eugene, OR 97403, USA

Eugene, OR 97403, USA

Eugene, OR 97403, USA

Email: dou@cs.uoregon.edu

Email: csehao@cs.uoregon.edu

Email: ahoyleo@cs.uoregon.edu

Abstract-Semantic Data Mining refers to the data mining tasks

that systematically incorporate domain knowledge, especially for
mal semantics, into the process. In the past, many research efforts
have attested the benefits of incorporating domain knowledge in
data mining. At the same time, the proliferation of knowledge
engineering has enriched the family of domain knowledge, espe
cially formal semantics and Semantic Web ontologies. Ontology is
an explicit specification of conceptualization and a formal way to
define the semantics of knowledge and data. The formal structure
of ontology makes it a nature way to encode domain knowledge
for the data mining use. In this survey paper, we introduce
general concepts of semantic data mining. We investigate why
ontology has the potential to help semantic data mining and
how formal semantics in ontologies can be incorporated into
the data mining process. We provide detail discussions for the
advances and state of art of ontology-based approaches and an
introduction of approaches that are based on other form of
knowledge representations.

I. INTRODU CTION
Data mmmg, also known as knowledge discovery from
database (KDD), is the process of nontrivial extraction of im
plicit, previously unknown, and potentially useful information
from data [30]. In the past few decades, advances in data
mining techniques lead to many remarkable revolutions in
data analytics and big data. Data mining also combines tech
niques from statistics, artificial intelligence, machine learning,
database system, and many other disciplines to analyze large
data sets. Semantic Data Mining refers to data mining tasks
that systematically incorporate domain knowledge, especially
formal semantics, into the process. The effectiveness of do
main knowledge in data mining has been attested in past
research efforts. Fayyad et al. [21] claimed that domain knowl
edge can play an important role in all stages of data mining
including, data transformation, feature reduction, algorithm
selection, post-processing, model interpretation and so forth.
Russell and Norvig [64] believed that an intelligent agent
(e.g., a data mining system) must have the ability to obtain
the background knowledge and should learn knowledge more
effectively with the background knowledge.
Previous semantic data mining research has attested the pos
itive influence of domain knowledge on data mining. For ex
ample, the preprocessing can benefit from domain knowledge
that can help filter out the redundant or inconsistent data [41],
[59]. During the searching and pattern generating process,

IEEE ICSC 2015,February 7-9,2015,Anaheim, California, USA

978-1-4799-7935-6115/$31.00 20 15 IEEE

domain knowledge can work as a set of prior knowledge of

constraints to help reduce search space and guide the search
path [8], [9]. Further more, the discovered patterns can be
cleaned out [49], [48] or made more visible by encoding them
in the formal structure of knowledge engineering [76].
To make use of domain knowledge in the data mining pro
cess, the first step must account for representing and building
the knowledge by models that the computer can further access
and process. The proliferation of knowledge engineering (KE)
has remarkably enriched the family of domain knowledge with
techniques that build and use domain knowledge in a fonnal
way [64]. Ontology is one of successful knowledge engineer
ing advances, which is the explicit specification of a concep
tualization [26], [67]. Normally, an ontology is developed to
specify a particular domain (e.g., genetics). Such an ontology,
often known as a domain ontology, formally specifies the
concepts and relationships in that domain. The encoded formal
semantics in ontologies is primarily used for effectively shar
ing and reusing of knowledge and data. Prominent examples
of domain ontologies include the Gene Ontology (GO [73]),
Unified Medical Language System (UMLS [45]), and more
than 300 ontologies in the National Center for Biomedical
Ontology (NCBO [2]).
Research in the area of the Semantic Web [10] has led to
quite mature standards for modeling and codifying domain
knowledge. Today, Semantic Web ontologies become a key
technology for intelligent knowledge processing, providing a
framework for sharing conceptual models about a domain. The
Web Ontology Language (OWL) [1], which has emerged as
the de facto standard for defining Semantic Web ontologies, is
widely used for this purpose. The Semantic Web technologies
that fonnally represent domain knowledge including structured
collection of prior infonnation, inference rules, knowledge
enriched datasets etc., could thus develop frameworks for
systematic incorporation of domain knowledge in an intelligent
data mining environment.
In this survey paper, we study the advances and state of
art of semantic data mining. We specifically focus on the
ontology-based approaches. The ontology-based approaches
for semantic data mining attempt to make use of formal on
tologies in the data mining process. This is generally achieved
by using the formal definition of concepts and relationships

244

Proceedings of the 20 IS IEEE 9th International Conference on Semantic Computing (IEEE ICSC 20 IS)
in ontologies as auxiliary information or constraint conditions
to guide the data mining process. For example, in classifi
cation, ontology can specify the consistency relationships of
the classification task. By ruling out the inconsistent search
space, the classification task would result in a better accu
racy [S]. Further more, structured organization of ontologies
can serve as a good representation for the data mining result.
For example, in information extraction and text mining, the
extracted information can be presented through the ontology
itself using an ontology definition language (e.g., OWL) [77].
In this paper, we focus on three perspectives of ontology-based
approaches in the research of semantic data mining:

Role of ontologies: Why domain knowledge with formal

semantics, such as ontologies, are necessary in all stages

of the data mining process.
Mining with ontologies: How ontologies are represented
and processed to help the data mining process.
Peiformance evaluation: How ontologies can improve the
performance of data mining systems in applications.

II. ROLE OF ONTOLOGIES IN SEMANTIC DATA MINING

The perspective and mechanism of utilizing ontologies in
semantic data mining varies across different systems and
applications. The question why ontology is useful in assisting
data mining process does not have an uniform conclusion.
By reviewing the previous ontology-based approaches, we
summarize the following three purposes that ontologies have
been introduced to semantic data mining:

To bridge the semantic gap between the data, applica

tions, data mining algorithms, and data mining results.
To provide data mining algorithms with a priori knowl
edge which either guides the mining process or re
duces/constrains the search space.
To provide a formal way for representing the data mining
flow, from data preprocessing to mining results.

A. Bridging the semantic gap

The question why domain knowledge is helpful in the data

mining process has been long discussed in previous semantic
data mining research. Researchers claim that there exists a
knowledge gap between the data, data mining algorithm,
and mining results in all stages of data mining including
preprocessing, algorithm execution, and result generation [ IS].
Data preprocessing usually contains data cleaning, normal
ization, transformation, feature extraction and selection. In
most scenarios, there exist semantic gaps in the steps of data
preprocessing. Without considering formal semantics, ad-hoc
or empirical methods are used to determine the quality of
the data. For example, scarcity and nearest neighbor rules are
usually adopt to determine the outliers and missing values.
In the normalization and transformation step, data semantics
is necessary for understanding the relations of the data. For
example, it is important to determine the correlation between
features and attributes of the data when performing data
normalization. Strongly correlated attributes could be reduced

into one combined attribute. In practice, semantic gaps are usu

ally filled manually by domain experts. However, ontologies
have been shown to be beneficial in many data preprocessing
tasks [41], [59], [72].
There exists semantic gap between the data mining algo
rithm and data as well. Data mining algorithms are usually
designed for data collected from different domains and sce
narios. However, data from a specific domain usually carry
domain specific semantics. The generic data mining algorithms
lack the ability to identify and make use of semantics across
different domains and applications. Ontologies are useful to
specify domain semantics and can reduce the semantic gap by
annotating the data with rich semantics. Semantic annotation
aims at assigning the basic element of information links to
formal semantic descriptions [20], [42]. Such elements should
constitute the semantics of their source. Semantic annotation
is crucial in realizing semantic data mining by bringing formal
semantics to data. The annotated data are very convenient for
the later steps of semantic data mining because the data are
promoted to the formal and structured format that connects
ontological terms and relations.
Many research efforts have dedicated to bridge the semantic
gap between data mining results and users. Marinica et al. [49],
[50] used ontology for the post pruning and filtering of the
association rule mining results. Mansingh et al. [4S] used on
tology to assist the subjective analysis for the association rule
post-pruning task. The data mining results can be represented
by ontologies in the semantic rich format which help sharing
and reuse. For example, information extraction (IE) is the task
of automatically extracting structured information from text.
The data/text mining results are sets of structured information
and knowledge with regarding to the domain. To represent
the structured and machine-readable information, it is nature
to represent the information with ontology. Ontology Based
Information Extraction (OBlE) [77] has extensively used this
representation. With OBIE, the information extracted is not
only well structured but also represented by predicates in the
ontology which are easy for sharing and reuse.
B. Providing prior knowledge and constraints

The definition and reuse of prior knowledge is one of the

most important problems for semantic data mining. As a
formal specification of concepts and relationships, ontology
is a nature way to encode the formal semantics of prior
knowledge. The encoded prior knowledge has the potential
to guide and influence all stages of the data mining process,
from preprocessing to result filtering and representation. For
example, Liu et al. [46] developed a RDF hypergraph represen
tation to capture information from both ontologies and data.
Ontologies are incorporated into the graph representation of
the data as the priori knowledge to bias the graph structure and
also representing the distances between terms and concepts
in the graph. The approach transforms the hypergraph and
weighted hyperedges into a bipartite graph to represent both
the data and ontology in a uniformed structure. Random
walks with restart over the bipartite graph is performed to

245

Proceedings of the 2015 IEEE 9th International Conference on Semantic Computing (IEEE ICSC 2015)
generate semantic associations. Whenever the random walk
goes through the ontology-based edges, the domain knowledge
encoded in ontologies bridges the latent semantic relations
underneath the data with rich semantics.
As a collection of concepts and predicates, ontology has the
ability to perform logic reasoning and thus make consistency
inference for those predicates. In semantic data mining, the
ability to make consistency inference is usually represented as
constraints. The set of constraints powered by the ontology
have the ability to detect inconsistent data and results in the
preprocessing stage, the algorithm execution stage, and the
result filtering and generation stage. For example, Balcan et
al. [8] incorporated ontology as consistency constraints into
multiple related classification tasks. The ontology specifies
the constraints between multiple classification tasks. Carlson
et al. [16] presented a semi-supervised information extraction
algorithm that couples the training of many information ex
tractors. Using ontology as constraints on the set of extractors,
it yields more accurate results. Claudia Marinica et al. [49],
[50] presented post-processing of the association rule mining
results using ontology for consistency checking. Invalid or
inconsistent association rules are pruned and filtered out with
the help of ontology and an inference engine.
C. Formally representing data mining results

The well designed data mining systems should present re

sults and discovered patterns in a formal and structured format,
so that data mining results are capable to be interpreted as
domain knowledge and to further enrich and improve current
knowledge bases. Ontology is one of the way to represent the
data mining results in a formal and structured way. As a formal
definition of concepts and relationships, ontology can encode
rich semantics for different domains. The data mining results
from different domains and tasks conform naturally with the
representation of ontology, for example, information extraction
and association rule mining. Specifically, in ontology-based
information extraction (aBlE) [55], [77], the extracted infor
mation are a set of annotated terms from the document with
the relations defined in the ontology. It is therefore straight
forward to represent the extracted information with ontology.
Wimalasuriya and Dou [77] claimed that ontology is a valid
form to represent the aBlE results in a semantic rich format.
Encoding aBlE results in the formal structure of ontology
could streamline the data mining process of other data mining
tasks that need to make use of the current result. The infer
ence engines which was designed in the field of knowledge
engineering could perform consistency checking that validate
the data mining results and clean out the inconsistent results.
aBlE systems can extract information with higher recall and
accuracy compared with traditional IE systems. The ontology
in aBlE provides the function as a conceptual framework and
consistency checking. It also organizes the extracted informa
tion in a formal and structured way using explicit ontology
representation. Similarly, ontology-based association pattern
mining method [46] can represent latent semantic associations.

III. MINING W ITH ONTOLOGIES

With formally encoded semantics, ontology has the potential
to assist in various data mining tasks. In this section, we sum
marize semantic data mining algorithms designed in several
important tasks, including association rule mining, classifica
tion, clustering, recommendation, information extraction, and
link prediction.
A. Ontology-based Association Rule Mining

Association rule mining is a fundamental data mining task

and well used in different applications. In the early work,
Svatek and Rauch [71] designed association mining tool that
can benefit from ontologies in all four stages of the mining
process: data understanding, task design, result interpretation,
and result dissemination over the Semantic Web. Bellandi et
al. [9] presented an ontology-based association rule mining
method, which queries the ontology to filter the instances used
in the association rule mining process. Ontology in this work
provides the constraints for queries in the association mining
process. The search space of association mining is constrained
by the query returned from the ontology that some items
from the output association rules are excluded or to be used
to characterize interesting items according to an abstraction
level. The user constraints include both pruning constraints,
which are used for filtering a set of non-interesting items, and
abstraction constraints, which permit a generalization of an
item to a concept of the ontology.
Marinica et al. [49], [50] presented post-processing of the
association rule mining results using an ontology for the
consistency checking. Invalid or inconsistent association rules
are pruned and filtered out with the help of ontology and
an inference engine. Recently, Liu et al. [46] proposed to
apply ontology and hypergraph to discover latent association
rules in the data. They built the connections between ontology
and data using a bipartite hypergraph model. Random walk
based metrics were proposed to measure the latent semantic
distances between concepts and terms. The term sets with
shorter semantic distances are ranked higher. Top ranked term
sets are generated as strong associations.
B. Ontology-based Classification

Classification is one of the most common data mining tasks

that finding a model (or function) to describe and distinguish
data classes or concepts [30]. In semantic data mining, one
typical use of ontology is to annotate the classification labels
with the set of relations defined in the ontology. Research
by Balcan et al. [8] indicates that with the ontology annotated
classification labels, the semantics encoded in the classification
task has the potential not only to influence the labeled data
in the classification task but also to handle large number of
unlabeled data. They incorporated ontology as consistency
constraints into multiple related classification tasks. These
tasks classify multiple categories in parallel. An ontology
specifies the constraints between the multiple classification
tasks. An unlabeled error rate is defined as the probability the
classifier assigns a label for the unlabeled data that violates the

246

Proceedings of the 2015 IEEE 9th International Conference on Semantic Computing (IEEE ICSC 2015)
ontology. This classification task produces the classification
hypothesis with the classifiers that produce the least unlabeled
error rate and thus most classification consistency.
Allahyari et al. [7] presented an ontology-based method for
automatic classification of text documents into a dynamically
defined set of topics of interest. Using DBpedia-based ontol
ogy, entities and relations among entities are identified from
the text document. Semantic graph of connected entities are
constructed from the set of relations. HITS algorithm [43] is
used to identify the core entities in the semantic graph for the
further identification of dynamic topics. The classification of
documents is based on calculating the similarity of document's
semantic graph to define ontological context (topics).
C. Ontology-based Clustering

Clustering [34] is a data mining task that grouping a set of

objects in the same cluster which are similar to each other.
Early work of ontology-based clustering includes using ontol
ogy in the text clustering task for the data preprocessing [31],
enriching term vectors with ontological concepts [32], and
promoting distance measure with ontology semantics [36].
Wei et al. [65] took advantage of the thesaurus-based and
corpus-based ontology for text clustering with the enriched
conceptual similarity. Song et al. [65] proposed a genetic
algorithm for text clustering with transformed latent semantic
indexing using ontology to capture the associated semantic
similarity. Jing et al. [35] used ontology to re-weight the
vectors in knowledge-based vector space for text clustering.
Fodeh [23] claimed that ontology can be used to greatly reduce
the number of features needed in the document clustering task.
With the aid of ontology, a core subset of semantic features
for each text corpus is identified. Using this core semantic
features for clustering, the number of features can be reduced
by 90% or more while still produce clusters that capture the
main themes in a text corpus.
Ovaska et al. [58] performed a gene clustering task from
microarray experiments with the aid of gene ontology (GO).
Graph structure (GS) and information content (IC) based mea
sures are used for the similarity measure between genes. GS
based methods use the hierarchical structure of GO to compute
the gene similarity. IC-based methods additionally consider
the information content of GO terms in a reference gene set.
Zhang et al. [82] proposed medical document clustering with
ontology-based term similarity measures. Ontology is used to
index the terms in the medical document set. The weight
of term is re-calculated by ontology-based term similarity
measure. Spherical K-mean is then used for the clustering task.
D. Ontology-based Information Extraction

Information extraction (IE) refers to the task of retrieving

certain types of information from natural language text by
processing them automatically. IE is closely related to text
mining. Ontology-based information extraction (OBIE) is a
subfield of information extraction, which uses formal ontolo
gies to guide the extraction process [40], [77]. Because of
this guidance in the extraction process, OBIE systems have

mostly implemented following a supervised approach [76].

Although very few semi-supervised IE systems are considered
as ontology-based [78], [79], they rely on instances of known
relationships [4], [63]. Therefore those semi-supervised sys
tems can also be considered as OBIE systems.
Early work of OBIE includes knowledge extraction from
web documents [5] and data-rich unstructured documents [19].
Ontology can provide consistency checking for the extracted
information in the IE system. Kara [39] presented an ontology
based information extraction and retrieval system which uses
ontology for consistency checking. The output of a regular
IE system is transformed to ontological instances through
ontology population. The inference and consistency check
ing are performed on these ontological instances. Carlson et
al. [17] proposed the semi-supervised information extraction
algorithm with few labeled data and large amount of unlabeled
data. The proposed algorithm couples multiple ontology based
information extractors with ontology specify the constraints
and exclusions for different categories and relations. The
algorithm iteratively and incrementally enrich the classification
label using most confident outputs of these extractors.
Recently, Fernandez et al. [22] presented a way to ex
ploit domain knowledge bases to support semantic search
capabilities in large document repositories. Nebhi [56] pro
posed an OBIE system for disambiguating Twitter messages.
By combing concepts from Freebase and extraction rules
based on dependency trees, Nebhi's approach determines the
meaning (and context) of entities mentioned in the messages.
Nebhi [57] improved the accuracy the disambiguation process
by replacing the pattern-based approach with a classification
task, using Support Vector Machine. As a way to promote
the adoption of OBIE, Wimalasuriya and Dou [76] proposed
the Ontology-based Components for Information Extraction
(OBCIE) architecture. OBCIE aims to encourage re-usability
by modeling the components of the IE system as modular
as possible. Gutierrez et al. [27] extended the OBCIE archi
tecture by incorporating hybrid configurations (e.g., different
implementations and different functionalities).
E. Ontology-based Recommendation System

Recommender systems or recommendation systems [3], [12]

are the systems that dedicate to predict the preference or
ratings that a user would give to an item. Recommendation
systems have become extremely popular in recent years and
been applied in a variety of applications including movies,
music, news, books, research articles, search queries, and
social tags [44], [75]. In a good rec Olmnendation system,
heterogeneous information from multiple sources is usually
required. Ontology can integrate the use of heterogeneous
information and guide the recommendation preference.
Early work of ontology-based recommendation system uses
ontology for user profiling [53], personalized search [60], and
web browsing [52], [51]. Recently, Pudota et al. [61] pro
posed a recommendation system that generate and recommend
tags automatically for web resources. The web documents
are annotated and matched by terms in the ontology first.

247

Proceedings of the 2015 IEEE 9th International Conference on Semantic Computing (IEEE ICSC 2015)
Then ontology-based reasoning is conducted to infer the new
knowledge from the annotated tenns. This inference is made
by finding the common ancestor nodes for them and possibly
all the nodes in the path between the matched nodes with on
tological concepts. Kang and Choi [38] proposed an ontology
based recommendation system in which the ontology is used
to encode the long term and short term preference information.
The user preference ontology is constructed from the concepts
of the general domain ontology together with the documents
that the user visited. Recommendation is made based on the
similarity between ontological concepts and terms.
IJntema [33] developed a recommendation system, Athena,
to provide ontology-based recommendation for the news feed
system. It extend the Herme framekwork [24], a framework
used to build a news personalization service, with the help of
ontology to detennine the semantic relations between tenns
and concepts. It uses an ontology to store concepts and
their relationships to the news items. Cantador et al. [14]
proposed another news recommendation system that makes use
of ontologies to provide online news recommendation services.
Domain ontologies are used to provide the concept framework
for news contents and user preferences. Domain ontologies can
automatically annotate news items with semantic concepts that
appear in both the textual contents and the domain ontologies.
F. Ontology-based Link P rediction

Link prediction for social networks becomes a very active

research area in data mining due to the success of online social
networks such as Twitter, Facebook, and Google+. Aljandal et
al. [6] presented a link prediction framework with ontology
enriched numerical graph features. The authors claimed that
in previous social network research flat representation of in
terest taxonomies limited the improvement of link prediction.
Ontology aggregated distance measure is proposed to encode
the interest taxonomies in ontology into the distance measure
to more accurately describe the shared user interests.
Thor et al. [74] presented a link prediction method on
ontology annotated data. The data are first annotated by
controlled vocabulary terms from ontologies. The annotation
links between the data and predicates in ontology fonn an
annotation graph. Graph summarization and dense subgraph
method were proposed to filter the graph and find promising
subgraphs. A scoring function based on multiple heuristics
was proposed to rank the predictions based on these filtered
subgraphs. Amakrishnan [62] proposed a method to discover
the informative connection subgraphs that relate two entities
in the graph. They proposed heuristics for edge weighting that
depend indirectly on the semantics of entity and property types
in the ontology and on characteristics of the instance data. The
display p-graph generation algorithm was proposed to extract
a small connection subgraph from the input graph. Mabroukeh
and Ezeife [47] proposed using the domain ontology for se
mantic web usage mining and next page prediction. Semantic
infonnation in the ontology is used in the sequential pattern
mining algorithm to prune the search space and partially
relieve the algorithm from support counting.

IV. PERF ORMANCE EVALUATION AND ApP LICATIONS

As a formal specification of domain concepts and relation
ships, ontology can assist in the data mining process in various
perspectives. It is reasonable to expect a performance gain
in ontology-based approaches compared with the data mining
approaches without using ontologies or other form of domain
knowledge. Many semantic data mining research efforts have
attested such improvements. With well designed algorithms,
previous research either reports performance improvement
or accomplishment of data mining tasks that could not be
achieved without using ontologies. In this section, we give
a brief smmnarization of the performance improvement in
ontology-based approaches and their applications.
A. Peiformance gain in precision, recall, and consistency of
data mining results

Many previous ontology-based efforts have reported per

formance gain in the data mining results. Ontology-based
approaches have been reported to have better precision and
recall than the traditional approaches in various tasks such
as text clustering [32], [33], [35], [65], [82], infonnation
extraction [17], [27], [56], [57], link prediction [6], [15], [74],
and rec Olmnendation systems [33], [52], [60], [61].
Research in recommendation system suggests that ontology
based recommendation systems have better prediction preci
sion than traditional recommendation methods [13], [83]. With
the enriched semantics and reduced search space, execution
speed gain has been reported in the gene clustering task from
microarray experiments with ontology-based clustering [58].
In the web usage mining and next page prediction task,
semantics-aware sequential pattern mining algorithms was
proved to perform 4 times faster than regular and non
semantics-aware algorithms [47].
Ontology-based approaches improve the consistency of data
mining results as well. Marinica et al. [49], [50] presented
post-processing of the association rule mining results using an
ontology for the consistency checking. Semantically inconsis
tent association rules are pruned and filtered out with the help
of ontology and logic reasoning.
B. Semantics rich data mining results

Ontology can also assist in enriching data mining results

with formal semantics. Semantics rich data mining results
are expected from ontology-based approaches compared with
approaches without using ontologies. For example, OBIE is
able to extract the information with similar or close semantics
that does not directly appear in the data source [39].
Without knowing semantics of the attributes or itemsets,
association rule mining usually generate too many rules or
even inconsistent rules. Ontology-based association rule min
ing bridges the semantic gap of the domain knowledge and the
association rule mining algorithm. It results in better collection
and representation of association rules by pruning the results
or reducing the search space. The top ranked rules also result
in high support measure for the targeting domain [9].

248

Proceedings of the 20 IS IEEE 9th International Conference on Semantic Computing (IEEE ICSC 20 IS)
With the aid of ontology, multi-level association rule mining
will discover concept-based association rules instead instance
based rules [29], [66], [80]. With supermarket transactions like
cheese and milk, bread and cake, etc., traditional association
rule mining methods have to generate rules with those items,
while multi-level association rule can generate conceptual
level rules like diary product -+ bakery products. The well
controlled granularity of semantics raises the potential of more
interesting association rules.
C. Peiforming data mining task that is unachievable with

traditional data mining methods

Certain data mining tasks that are not achievable by tradi

tional data mining methods can be accomplished by ontology
based approaches. For example, traditional classification task
usually requires at least reasonable amount of labeled data as
prior knowledge. Using ontology as the specification of prior
knowledge, classification task without enough labeled data
is proved to have a comparable performance compared with
traditional classification methods [8]. Using the ontology as
a conceptual consistency constraint, the model with unlabeled
data can be tuned into the one that have the best consistency
with the prior knowledge (i.e., ontology). Classification task
without labeled or annotated data is also reported in the
ontology-based text classification task [7].
V. OTH ER ApPROACH ES IN SEMANTIC DATA MINING
Although ontology is one of the most common ways for
formally representing domain knowledge, other representa
tions of domain knowledge have been used in semantic data
mining. Early research efforts of semantic data mining have
employed concept hierarchy as a very important representa
tion of domain knowledge. Previous concept hierarchy based
algorithms largely focus on exploiting its generalization ability
that it could handle the raw data at higher conceptual level.
Han et al. [28] use concept hierarchy to guide such general
ization process of attributes in quantitative association mining.
Later, Han and Fu [29] proposed multi-level association rule
mining using concept hierarchy to control the granularity of
knowledge discovered from data at different conceptual levels.
Kamber et al. [37] proposed concept hierarchy based decision
tree methods in which the induction of decision trees could
be achieved at different levels of conceptual abstraction.
Later, some research efforts have employed knowledge
bases for semantic data mining tasks, including Wikipedia
and Freebase [11], which are not exactly formal ontologies.
Gabrilovich and Markovitch [25] computed the semantic re
latedness using Wikipedia-based semantic analysis in which
substantial improvements in computing word and text related
ness is confirmed. Milne and Witten [54] deployed Wikipedia
as the external knowledge for the document clustering task.
Significant performance improvement has been achieved using
concept and category information in Wikipedia to annotate the
documents with enriched semantics information. Yu et al. [81]
explored a way to build personalized entity recommendation
framework for search engine users by utilizing the knowledge

extracted from Freebase. A user log dataset collected from a

commercial search engine together with the entity graph ex
tracted from Freebase are used to generate semantic enriched
features and build up recommendation models.
Most recently, a new representation, Meta-path, has been
designed for semantic data mining tasks. The meta-path is a
path that defines a composition of relations between the set of
terms on the path [69]. For example, in the bibliographic net
work, a typical meta-path could be author -+ paper -+ venue.
It is usually defined based on the graph of network schema
of related data mining terms and concepts. Comparing with
an ontology, each meta-path could relate to multiple concepts
while each predicate in an OWL ontology usually is related
to two concepts. The type of meta-path is defined by the type
of entities in the meta-path while the type of predicate in an
ontology is defined by the related concepts. Recent research
efforts on meta-path have successfully shown its capability to
explore efficient semantic data mining algorithm from many
perspective. Early work of meta-path focuses on exploiting the
semantic enriched similarity representation in heterogenous
information network on various applications [68], [70]. Sun et
al. [68] presented PathSim, a meta-path based similarity search
method in which the semantic similarities between entities
are calculated according to the structure of meta-path and the
matrix representation of data instance relations. Sun et al. [70]
also proposed PathSelClus, a meta-path based clustering algo
rithm in which semantic similarities are measured based on a
probabilistic model on the meta-path framework.
V I. CONCLUS ION
The advances in knowledge engineering and data mining
promote semantic data mining, which brings rich semantics to
all stages of data mining process. Many research efforts have
attested the advantage of incorporating domain knowledge
into data mining. Formal semantics encoded in the ontology
is well structured which is easy for the machine to read
and process thus make it a nature way to use ontologies in
semantic data mining. Using ontologies, semantic data mining
has advantages to bridge semantic gaps between the data,
applications, data mining algorithms, and data mining results,
provide the data mining algorithm with priori knowledge
which either guides the mining process or reduces the search
space, and to provide a formal way for representing the data
mining flow, from data preprocessing to mining results.
In the past decade, to handle and manipulate the big data
have raised intense discussion in the data mining community.
With the development of knowledge engineering, especially
Semantic Web techniques, mining large amount, semantics
rich, and heterogeneous data emerges as an important research
topic in the community. As many researchers have pointed out,
work along semantic data mining is still in its early stage.
Ontology-based semantic data mining seems to be one of
most promising approaches. The major challenge is to develop
more automatic semantic data mining algorithms and systems
by utilizing the full strength of formal ontology that has

249

Proceedings of the 20 IS IEEE 9th International Conference on Semantic Computing (IEEE ICSC 20 IS)
well defined representation language, formal semantics, and
reasoning tools for logic inference and consistency checking.
ACKNOW LEDGMENT

We would like to thank the paper invitation from the orga

nizing committee of the 2015 IEEE International Conference
on Semantic Computing. This research has been partially
supported by the NSF grant IIS-1118050 and NIH grant
ROIGM103309. The views and conclusions contained in this
document are those of the authors and should not be inter
preted as representing the official policies, either expressed or
implied, of the NSF, the NIH, or the U.S. Government.
REF ERENCES
[I] OWL Web Ontology Language.
http://www.w3.org/TR/owl-ref/.
[2] The National Center for Biomedical Ontology.
http://www.bioontology.org/.
[3] G. Adomavicius and A. Tuzhilin. Toward the next generation of
recommender systems: A survey of the state-of-the-art and possible
extensions. Knowledge and Data Engineering, IEEE Transactions on,
17(6):734-749,2005.
[4] E. Agichtein and L. Gravano. S nowball: Extracting relations from large
plain-text collections. In Proceedings of the 5th ACM International
Conference on Digital Libraries, pages 85-94,2000.
[5] H. Alani, S . Kim, D. E. Millard, M. J. Weal, W. H all, P. H. Lewis, and
N. R. S hadbolt. Automatic ontology-based knowledge extraction from
web documents. Intelligent Systems, IEEE, 18(1):14-21,2003.
[6] W. Aljandal, Y. Bahirwani, D. Caragea, and W. H. H su. Ontology-aware
classification and association rule mining for interest and link prediction
in social networks. In AMI Spring Symposium: Social Semantic Web:
Where Web 2.0 Meets Web 3.0, pages 3-8,2009.
[7] M. Allahyari, K. J. Kochut, and M. Janik.
Ontology-based text
classification into dynamically defined topics. In Semantic Computing
(ICSC), 2014 IEEE International Conference on, pages 273-278,2014.
[8] N. Balcan, A. Blum, and Y. Mansour. Exploiting ontology structures
and unlabeled data for learning. In Proceedings of the 30th International
Conference on Machine Learning, pages 1112-1120,2013.
[9] A. Bellandi, B. F urletti, Y. Grossi, and A. Romei. Ontology-driven
association rule extraction: A case study. Contexts and Ontologies
Representation and Reasoning, page 10,2007.
[10] T. Bemers-Lee, J. H endler, and O. Lassila. The semantic web. Scientific
American, 284(5):28-37,2001.
[11] K. BoUacker, C. Evans, P. Paritosh, T. S turge, and 1. Taylor. F reebase: a
collaboratively created graph database for structuring human knowledge.
In Proceedings of the 2008 ACM SIGMOD international conference on
Management of data, pages 1247-1250,2008.
[12] R. Burke. H ybrid recommender systems: S urvey and experiments. User
modeling and user-adapted interaction, 12(4):331-370,2002.
[13] I. Cantador, A. Bellogin, and P. Castells. A multilayer ontology-based
hybrid recommendation model. Al Communications, 21(2):203-210,
2008.
[14] I. Cantador, A. Bellogin, and P. Castells. Ontology-based personalised
and context-aware recommendations of news items. In Proceedings of
the 2008 IEEEIWICIACM International Conference on Web Intelligence

pages 562-565,2008.
[l5] D. Caragea, Y. Bahirwani, W. Aljandal, and W. H. H su. Ontology-based
link prediction in the livejournal social network. In SARA, volume 9,
pages 1-1,2009.
[16] A. Carlson, 1. Betteridge, E. R. H ruschka Jr, and T. M. Mitchell.
Coupling semi-supervised learning of categories and relations. In
and Intelligent Agent Technology-Volume 01,

Proceedings of the NAACL HLT 2009 Workshop on Semi-Supervised

Learning for Natural Language Processing, pages 1-9,2009.
[17] A. Carlson, J. Betteridge, R. C. Wang, E. R. H ruschka Jr, and T. M.
Mitchell. Coupled semi-supervised learning for information extraction.
In Proceedings of the third ACM international conference on Web Search
and Data Mining, pages 101-110, 2010.
[18] P. Domingos. Toward knowledge-rich data mining. Data Mining and
Knowledge Discovery, 15(1):21-28,2007.

[19] D. W. Embley, D. M. Campbell, R. D. S mith, and S . W. Liddle.

Ontology-based extraction and structuring of information from data-rich
unstructured documents. In Proceedings of the seventh international
conference on Information and knowledge management, pages 52-59.
ACM, 1998.
[20] M. Erdmann, A. Maedche, H .-P. S chnurr, and S . S taab. F rom manual to
semi-automatic semantic annotation: About ontology-based text annota
tion tools. In Proceedings of the COLlNG-2000 Workshop on Semantic
Annotation and Intelligent Content, pages 79-85,2000.
[21] U. F ayyad, G. P iatetsky-S hapiro, and P. S myth. F rom data mining to
knowledge discovery in databases. Al magazine, 17(3):37,1996.
[22] M. F ernandez, l. Cantador, Y. Lopez, D. Vallet, P. Castells, and
E. Motta. S emantically enhanced information retrieval: an ontology
based approach. Web Semantics: Science, Services and Agents on the
World Wide Web, 9(4):434-452,2011.
[23] S . F odeh, B. Punch, and P.-N. Tan. On ontology-driven document
clustering using core semantic features. Knowledge and information
systems, 28(2):395-421,2011.
[24] F. F rasincar, J. Borsje, and L. Levering. A semantic web-based approach
for building personalized news services. International Journal of E
Business Research (!JEBR), 5(3):35-53,2009.
[25] E. Gabrilovich and S . Markovitch. Computing semantic relatedness
using wikipedia-based explicit semantic analysis. In !JCAl, volume 7,
pages 1606-1611,2007.
[26] T. R. Gruber. Toward principles for the design of ontologies used for
knowledge sharing? International journal of human-computer studies,
43(5):907-928,1995.
[27] F. Gutierrez, D. Dou, A. Martini, S . F ickas, and H. Zong. H ybrid
ontology-based information extraction for automated text grading. In
Machine Learning and Applications (ICMLA), 2013 12th International
Conference on, volume 1, pages 359-364. IEEE, 2013.
[28] J. H an, Y. Cai, and N. Cercone. Data-driven discovery of quantitative
rules in relational databases. Knowledge and Data Engineering, IEEE
Transactions on, 5(1):29-40,1993.
[29] J. H an and Y. F u. Discovery of multiple-level association rules from
large databases. In VLDB, volume 95, pages 420-431,1995.
[30] J. H an and M. Kamber. Data Mining: Concepts and Techniques. Morgan
Kaufmann Publishers Inc., S an F rancisco, CA, US A, 2005.
[31] A. H otho, A. Maedche, and S . S taab. Ontology-based text document
clustering. KI, 16(4):48-54,2002.
[32] A. H otho, S . S taab, and G. S tumme. Ontologies improve text document
clustering. In Proceedings of the third IEEE International Conference
on Data Mining, pages 541-544,2003.
[33] w. IJntema, F. Goossen, F. F rasincar, and F. H ogenboom. Ontology
based news recommendation. In Proceedings of the 2010 EDBT/ICDT
Workshops, page 16. ACM, 2010.
[34] A. K. Jain. Data clustering: 50 years beyond k-means. Pattern
Recognition Letters, 31(8):651-666,2010.
[35] L. Jing, M. K. Ng, and J. Z. H uang. Knowledge-based vector space
model for text clustering. Knowledge and information systems, 25(1):3555,2010.
[36] L. Jing, L. Zhou, M. K. Ng, and J. Z. H uang. Ontology-based distance
measure for text clustering. In Proc. of SIAM SDM workshop on text
mining, Bethesda, Maryland, USA, 2006.
[37] M. Kamber, L. Winstone, W. Gong, S . Cheng, and J. H an. Generalization
and decision tree induction: Efficient classification in data mining.
In Research Issues in Data Engineering, 1997. Proceedings. Seventh
International Workshop on, pages 111-120. IEEE, 1997.
[38] J. Kang and J. Choi. An ontology-based recommendation system
using long-term and short-term preferences. In Information Science
and Applications (IC/SA), 2011 International Conference on, pages 1-8.
IEEE, 2011.
[39] S . Kara, O. Alan, O. S abuncu, S . Akpmar, N. K. Cicekli, and F. N.
Alpaslan. An ontology-based retrieval system using semantic indexing.
Information Systems, 37(4):294-305, 2012.
[40] Y. Karkaletsis, P. F ragkou, G. Petasis, and E. Iosif. Ontology based
information extraction from text. In G. Paliouras, C. S pyropoulos,
and G. Tsatsaronis, editors, Knowledge-Driven Multimedia Information
Extraction and Ontology Evolution, volume 6050 of Lecture Notes in
Computer Science, pages 89-109. S pringer Berlin H eidelberg, 2011.
[41] N. Khasawneh and c.-c. Chan. Active user-based and ontology-based
web log data preprocessing for web usage mining. In Proceedings of

250

the 2006IEEEIWIClACM International Conference on Web Intelligence,

pages 325-328,2006.

Proceedings of the 2015 IEEE 9th International Conference on Semantic Computing (IEEE ICSC 2015)
[42] A. Kiryakov, 8 . P opov, I. Terziev, D. Manov, and D. Ognyanoff.
S emantic annotation, indexing, and retrieval. Web Semantics: Science,
Services and Agents on the World Wide Web, 2(1):49-79,2004.
[43] J. M. Kleinberg, R. Kumar, P. Raghavan, S . Rajagopalan, and A. S .
Tomkins. The web as a graph: Measurements, models, and methods. In
Computing and combinatorics, pages 1-l7. S pringer, 1999.
[44] Y. Koren. The beUkor solution to the netfl ix grand prize. Nerjlix prize
documentation, 81,2009.
[45] D. Lindberg, B. H umphries, and A. McCray. The Unified Medical
Language S ystem. Methods of Information in Medicine, 32(4):281-291,
1993.
[46] H. Liu, D. Dou, R. Jin, P. LePendu, and N. S hah. Mining biomedical
ontologies and data using rdf hypergraphs. In Machine Learning
and Applications (ICMLA), 2013 12th International Conference on,

volume I , pages 141-146. IEEE, 2013.

[65] W. S ong, C. H. Li, and S . C. Park. Genetic algorithm for text clustering
using ontology and evaluating the validity of various semantic similarity
measures. Expert Systems with Applications, 36(5):9095-9104,2009.
[66] R. S rikant and R. Agrawal. Mining generalized association rules. In
VLDB, volume 95, pages 407-419,1995.
[67] R. S tuder, V. R. Benjamins, and D. F ensel. Knowledge engineering:
principles and methods. Data & knowledge engineering, 25(1):161197,1998.
[68] Y. S un, J. H an, X. Yan, P. S . Yu, and T. Wu. Pathsim: Meta path
based top-k similarity search in heterogeneous information networks. In
V LDB ' l 1, pages 992-1003,2011.
[69] Y. S un, B. Norick, J. H an, X. Yan, P. S . Yu, and X. Yu. Integrating
meta-path selection with user-guided object clustering in heterogeneous
information networks. In Proceedings of the 18th ACM SIGKDD
international conference on Knowledge discovery and data mining,

Proceedings of the 1st international conference on Knowledge capture,

pages 1348-1356,2012.
[70] Y. S un, B. Norick, J. H an, X. Yan, P. S . Yu, and X. Yu. Pathselclus:
Integrating meta-path selection with user-guided object clustering in
heterogeneous information networks. ACM Transactions on Knowledge
Discovery from Data (TKDD), 7(3):11,2013.
[71] V. S vatek, J. Rauch, and M. RalbovskY . Ontology-enhanced association
mining. In Semantics, Web and Mining, pages 163-179. S pringer, 2006.
[72] D. Tanasa and B. Trousse. Advanced data preprocessing for intersites
web usage mining. Intelligent Systems, IEEE, 19(2):59-65,2004.
[73] The ene_ontology_ consortium. Creating the gene ontology resource:
design and implementation. Genome Res., 11(8):1425-1433, August
2001.
[74] A. Thor, P. Anderson, L. Raschid, S . Navlakha, B. S aha, S . Khuller,
and X .-N. Zhang. Link prediction for annotation graphs using graph
summarization. In T he Semantic Web-ISW C 2011, pages 714-729.
S pringer, 2011.
[75] A. Toscher, M. Jahrer, and R. M. Bell. The bigchaos solution to the
nettl ix grand prize. Nerjlix prize documentation, 2009.
[76] D. C. Wimalasuriya and D. Dou. Components for information extraction:
Ontology-based information extractors and generic platforms. In Pro

pages 100-107. ACM, 2001.

[53] S . E. Middleton, N. R. S hadbolt, and D. C. De Roure. Ontological user
profiling in recommender systems. ACM Transactions on Information
Systems (TOIS), 22(1):54-88,2004.
[54] D. Milne and I. H. Witten. An effective, low-cost measure of semantic
relatedness obtained from wikipedia links. In Proceeding of AAAI

pages 9-18,2010.
[77] D. C. Wimalasuriya and D. Dou. Ontology-based information extraction:
An introduction and a survey of current approaches. Journal of
Information Science, 36(3):306-323,2010.
[78] F. Wu and D. S . Weld. Autonomously semantifying wikipedia. In Pro

[47] N. R. Mabroukeh and C. I. Ezeife. Using domain ontology for semantic

web usage mining and next page prediction. In Proceedings of the J 8th
ACM conference on Information and knowledge management, pages
1677-1680. ACM, 2009.
[48] G. Mansingh, K -M. Osei-Bryson, and H. Reichgelt. Using ontologies
to facilitate post-processing of association rules by domain experts.
Information Sciences, 181(3):419-434,2011.
[49] c. Marinica and F. Guillet. Knowledge-based interactive postmining of
association rules using ontologies. Knowledge and Data Engineering,
IEEE Transactions on, 22(6):784-797,2010.
[50] c. Marinica, F. Guillet, and H. Briand. P ost-processing of discovered
association rules using ontologies. In Data Mining Workshops, 2008.
ICDMW '08. IEEE International Conference on, pages 126-133,2008.
[51] S . E. Middleton, H. Alani, and D. C. De Roure. Exploiting synergy be
tween ontologies and recommender systems. arXiv preprint csI0204012,
2002.
[52] S . E. Middleton, D. C. De Roure, and N. R. S hadbolt. Capturing
knowledge of user preferences: ontologies in recommender systems. In

Workshop on Wikipedia and Artificial Intelligence: an Evolving Synergy,

pages 25-30,2008.
[55] H .-M. MU lI er, E. E. Kenny, and P. W. S ternberg. Textpresso: an
ontology-based information retrieval and extraction system for biological
literature. PLoS biology, 2(11):e309,2004.
[56] K Nebhi. Ontology-based information extraction from twitter. In
Workshop on Information Extraction and Entity Analytics on Social

pages l7-22,2012.
[57] K Nebhi. Named entity disambiguation using freebase and syntactic
parsing. In LD4IE @ IS W C, 2013.
[58] K. Ovaska, M. Laakso, and S . H autaniemi. F ast gene ontology based
clustering for microarray experiments. BioData mining, 1(1):11,2008.
[59] D. Perez-Rey, A. Anguita, and 1. Crespo. Ontodataclean: Ontology
based integration and preprocessing of distributed data. In Biological
and Medical Data Analysis, pages 262-272. S pringer, 2006.
[60] A. P retschner and S . Gauch. Ontology based personalized search. In
Media Data - COLING 2012,

Proceedings of the 11th IEEE International Conference on Tools with

Artificial Intelligence, pages 391-398,1999.
[61] N. Pudota, A. Dattolo, A. Baruzzo, F. F errara, and C. Tasso. Automatic
keyphrase extraction and ontology mining for content-based tag recom
mendation. International Journal of Intelligent Systems, 25(12):11581186,2010.
[62] C. Ramakrishnan, W. H. Milnor, M. P erry, and A. P. S heth. Discovering
informative connection subgraphs in multi-relational graphs. ACM
SIGKDD Explorations Newsletter, 7(2):56-63,2005.
[63] S . Riedel, L. Yao, and A. McCallum. Modeling relations and their
mentions without labeled text. In Proceedings of the 2010 European

ceedings of the 19th ACM Conference on Information and Knowledge

Management (CIKM),

ceedings of the 16th ACM Conference on Information and Knowledge

Management, CIKM ' 07, pages 41-50,2007.
[79] F. Wu and D. S . Weld. Automatically refining the wikipedia infobox
ontology. In Proceedings of the 17th international conference on World
Wide Web, WWW ' 08, pages 635-644. ACM, 2008.
[80] G. Yang, K S himada, S . Mabu, K H irasawa, and J. H u. A genetic
network programming based method to mine generalized association
rules with ontology. In SICE, 2007 Annual Conference, pages 27152722. IEEE, 2007.
[81] X. Yu, H. Ma, 8 .-J. P. H su, and J. H an. On building entity recommender
systems using user click log and freebase knowledge. In Proceedings of
the 7th ACM international conference on Web search and data mining,

pages 263-272,2014.
[82] X. Zhang, L. Jing, X. H u, M. Ng, 1. X ia, and X. Zhou. Medical document
clustering using ontology-based term similarity measures. International
Journal of Data Warehousing and Mining (IJDWM), 4(1):62-73,2008.
[83] L. Zhuhadar, O. Nasraoui, R. Wyatt, and E. Romero. Multi-model
ontology-based hybrid recommender system in e-I earning domain. In

conference on Machine learning and knowledge discovery in databases:

TTl, ECML PKDD' I O , pages 148-163, Berlin, H eidelberg, 2010.

S pringer-Verlag.
[64] S . 1. Russell and P. Norvig. Artificial Intelligence: A Modern Approach.
Pearson Education, 2 edition, 2003.
Part

251

Proceedings of the 2009 1EEEIWICIACM International Joint Conference

on Web Intelligence and Intelligent Agent Technology-Volume 03,

91-95,2009.

pages